Identification of time-correlated neutrino clusters in populations of astrophysical transient sources

Mathieu Lamoureux,Gwenhaël de Wasseige

DOI: https://doi.org/10.22323/1.444.1507

2023-08-22

Abstract:The detection of astrophysical neutrinos from transient sources can help to understand the origin of the neutrino diffuse flux and to constrain the underlying production mechanisms. In particular, proton-neutron collisions may produce GeV neutrinos. However, at these energies, neutrino data from large water Cherenkov telescopes, like KM3NeT and IceCube, are dominated by the well-known atmospheric neutrino flux. It is then necessary to identify a sub-dominant component due to an astrophysical emission based on time correlation across messengers. The contribution covers several methods to search for such a signal in short time windows centered on observed transient sources, including a novel approach based on the distribution of time differences. Their performance is compared in the context of subpopulations of astrophysical sources that may show prompt or delayed neutrino emissions. The outlook for the usage of such techniques in actual analyses is also presented.

Instrumentation and Methods for Astrophysics,High Energy Astrophysical Phenomena

What problem does this paper attempt to address?

This paper aims to solve the problem of how to identify time - related neutrino clusters from transient astrophysical sources. Specifically, the research objective is to distinguish the atmospheric neutrino background from the potential astrophysical neutrino signal through time correlation, so as to better understand the origin of the neutrino diffuse flux and constrain its production mechanism.

### Main problems:
1. **Interference of background noise**: In the GeV energy range, neutrino data from large - scale water - Cherenkov telescopes (such as KM3NeT and IceCube) are mainly dominated by the known atmospheric neutrino flux. Therefore, it becomes very difficult to identify the sub - dominant astrophysical emission signals.
2. **Utilization of time correlation**: For transient sources, their neutrino emissions may have a certain delay in time compared with other messengers (such as gamma - rays). The specific value of this delay is unknown, but it may be consistent for a certain type of source or a certain emission mechanism. Therefore, by looking for neutrino excesses with the same delay, the influence of background events can be reduced.

### Solutions:
The paper proposes several methods to search for neutrino signals within these short - time windows:

1. **Maximum Likelihood Estimation (MLE) method**:
- By defining the likelihood function \( L(N, \{t_i\}; n_s, \Delta t, \sigma_t) \), where \( N \) is the total number of events, \( \{t_i\} \) is the detection time, \( n_s \) is the expected number of signal events, \( \Delta t \) and \( \sigma_t \) are the signal central time and width respectively, and \( S \) and \( B \) are the probability density functions (PDFs) of the signal and background respectively.
- Use the maximum likelihood estimation (MLE) to evaluate the existence and significance of the signal.

2. **Principal Component Analysis (PCA) method**:
- Use the PCA dimension - reduction technique to convert the observed relative times \( \{t_i\} \) into low - dimensional features, so as to make cluster analysis easier.
- Select an appropriate threshold to distinguish background events from potentially interesting time windows.

3. **PeANuTS method**:
- This method is based on the Poisson - expectation Anomaly for Neutrino Transient Source Search. It identifies events that deviate from the Poisson distribution by calculating the test statistic \( TS_{m,i} \) for each event.
- The calculation formula of \( TS_{m,i} \) is:
\[
TS_{m,i} = -\sum_{\Delta t \in D_{m,i}} \log \left[ 1 - \exp(-r_{\text{bkg}} \Delta t) \right]
\]
- The test statistic \( TS \) follows the Erlang distribution, and interesting events can be selected by setting an appropriate threshold \( TS_{\text{thr}} \).

### Results:
- Simulation experiments in different scenarios show that the PeANuTS method is superior to the MLE and PCA methods in most cases, especially when dealing with multiple signals.
- The PeANuTS method can significantly reduce the signal strength threshold and improve the detection ability of weak signals.

### Summary:
This paper proposes a new method - PeANuTS, for identifying time - related neutrino clusters in transient astrophysical sources. By using the correlation of signal times, this method can more effectively identify potential astrophysical neutrino signals in the case of high background noise.